Detection of variation in DNA sequences forms the basis of many applications in modern genetic analysis: it is used in linkage analysis to track disease genes in human pedigrees or economically important traits in animal and plant breeding programmes; it forms the basis of fingerprinting methods used in forensic and paternity testing [Krawczak and Schmidtke, 1994]; it is used to discover mutations in biologically and clinically important genes [Cooper and Krawczak, 1989]. The importance of DNA polymorphism is underlined by the large number of methods that have been developed to detect and measure it [Cotton, 1993]. Most of these methods depend on one of two analytical procedures, gel electrophoresis or molecular reassociation, to detect sequence variation. Each of these powerful procedures has its drawbacks. Gel electrophoresis has very high resolving power, and is especially useful for the detection of variation in the mini- and microsatellite markers that are used in linkage analysis and fingerprinting; it is also the method used to analyse the variation found in the triplet repeats that cause a number of mutations now known to be the cause of around ten genetic disorders in humans [Willems, 1994]. Despite its great success and widespread use, gel electrophoresis has proved difficult to automate: even the systems which automate data collection require manual gel preparation; and as samples are loaded by hand, it is easy to confuse samples. The continuous reading electrophoresis machines are expensive, and manual analysis is technically demanding, so that its use is confined to specialised laboratories which have a high throughput. Furthermore, difficulties in measuring fragment size preclude rigorous statistical analysis of the results.
By contrast, oligonucleotide hybridisation lends itself to automation and to quantitative analysis [Southern et al., 1992], but it is not well suited to the analysis of variation in the number of repeats in the micro- and minisatellites, as the small fractional change in the number of repeats produces a barely detectable change in signal strength; and of course it would not be possible to distinguish two alleles in the same sample as each would contribute to a single intensity measurement. Thus, many different combinations of alleles would produce the same signal. Present hybridisation methods are much better suited to analysing variation in the DNA due to point mutation--base substitution deletions and insertions, for which it is possible to design allele specific oligonucleotides (ASOs) that recognise both the wild type and the mutant sequences [Conner et al., 1983]. Thus it is possible in principle, in a relatively simple test, to detect all possible genotypes. However, a problem that arises in practice in the use of oligonucleotide hybridisation is that in some cases the extent of reassociation is barely affected by a mismatched base pair.